Recognizing the activities performed by the user in an unobtrusive manner is one of the important requisites of pervasive computing. Users perform a number of activities during their day to day life. Tracking and deciding what a user is doing at a given time involves a number of challenges. The lack of a precise pattern in doing an activity at different times is one among them. The number, order, and duration of the different steps involved in an activity vary significantly, even when the activity is done by the same user at different times. To overcome these challenges, a number of simultaneous inputs have to be handled with provisions for handling variations in number, order and duration of these inputs. This paper explains how multi-tape fuzzy finite state automata can be used to effectively recognize human activities. The method explained is found to give good results when tested using publicly available activity datasets collected in a smart home environment.